An AI-powered app that turns natural language into SQL queries
Easily query your databases by simply asking questions in plain English. Supports multiple databases, seamless query execution, and a clean interface for instant insights. App Link
- 🔹 Convert natural language → SQL instantly
- 🔹 Works with PostgreSQL, MySQL, SQLite, and more
- 🔹 User-friendly UI built with Streamlit
- 🔹 Powered by LLMs for accurate query generation
- 🔹 Preview and run SQL queries directly on your data
- 🔹 Can be extended to domain-specific chatbots (e.g., food, e-commerce, finance)
👨💻 Perfect for analysts, developers, product teams, and businesses who want to make data access as simple as asking a question.
This project uses the Zomato restaurants dataset from Kaggle to demonstrate how non-technical users can ask simple queries and receive:
- Relevant SQL queries generated automatically
- Meaningful recommendations in plain English
- Direct links to ordering platforms or restaurant pages
In other words: it’s like having a personal data assistant for restaurants. Users can ask:
“Where can I get biryani near me at 11 pm?”
and instantly get the best options available.
Prompt2SQL can be adapted to solve multiple real-world problems:
-
Customer Support Chatbots
- Transform user queries into database lookups (e.g., order status, product availability).
- Reduces dependency on customer care agents.
-
Food Delivery & Restaurant Discovery
- Users can ask for cuisines, timings, ratings, or locations in natural language.
- Personalized recommendations without complex filtering.
-
E-commerce Search & Recommendation
- Shoppers can type “Show me red sneakers under ₹3000 with 4+ stars”.
- SQL is generated, fetching results directly from the catalog database.
-
Business Intelligence (BI) & Analytics
- Non-technical managers can ask: “Show last quarter revenue by region”.
- Eliminates the need to write SQL or wait for analysts.
-
Finance & Banking
- Customers ask: “Show my last 5 transactions over ₹10,000”.
- SQL pulls data securely from their banking database.
Q: Want to have some biryani for dinner, show some places open at 11pm
A:
[('Hotel Hyderabad Chicken & Beef Biryani House', 'https://www.zomato.com/hyderabad/hotel-hyderabad-chicken-beef-biryani-house-chanda-nagar/info'),
('The Village Food Court', 'https://www.zomato.com/hyderabad/the-village-food-court-suraram/info'),
('Dabbawala Biryani', 'https://www.zomato.com/hyderabad/dabbawala-biryani-madhapur/info'),
... ]
Q: Currently I'm in Park Street and want to have the best Chinese dishes, show some options
A:
[('Mamagoto', 'Excellent', '4.7', 'https://www.zomato.com/kolkata/mamagoto-park-street-area/info'),
('BarBQ', 'Very Good', '4.4', 'https://www.zomato.com/kolkata/barbq-park-street-area/info'),
('Aaira', 'Very Good', '4.4', 'https://www.zomato.com/kolkata/aaira-park-street-area/info'),
... ]
Install dependencies:
pip install -r requirements.txt
Get a Gemini API Key → Gemini Pro
Set it as an environment variable:
GOOGLE_API_KEY="your_api_key"
For local development, I first used MySQL with a Streamlit app on localhost.
For deployment, I switched to Neon DB as a cloud-based database alternative.
- Add support for schema understanding & auto-joins
- Deploy as FastAPI + Docker microservice
- Add authentication & role-based query access
- Extend to finance / e-commerce datasets